Mac ARMs challenge skeptics: First tests of the Mac mini for developers leave Surface Pro X biting the dust
The transition from Macs to "Apple Silicon" represents a radical change for those computers, which leave Intel's chips behind. The change raises many questions, but one is obvious: how powerful will those chips and Macs be?
Now we begin to catch a glimpse of the answer: Early performance tests from developer kits perform better than Surface Pro Xs with Qualcomm's more powerful chips, and they do so with a major downside.
A performance that stands out and that above all promises
Several developers have used the well-known Geekbench 5 benchmark to publish the results of the Apple Developer Transition Kit (DTK) in this test.
So the DTK with a two year old iPad chip runs x86_64 code, in emulation, faster than the Surface Pro X runs it natively 😅 Oh boy Qualcomm, what are you even doing? https://t.co/UAlZiwSsF8- Steve Troughton-Smith (@stroughtonsmith) June 29, 2020
This "Mac mini ARM" is a team that Apple has begun to send developers to move their applications to new processors and thus prepare for the transition. It is a team with an Apple A12Z SoC, 16 GB of RAM and 512 GB of SSD and that has macOS Big Sur pre-installed in its developer version and that also has the Xcode development platform.
The Geekbench 5 scores of these teams are remarkable: taking the average of several published results, the team achieved 811 points in Single Core and 2,871 points in Multi Core in that test. A Surface pro X based on the Qualcomm SQ1, the firm's most powerful chip - which runs at 3 GHz - for convertibles based on Windows 10, achieves 726 in Single Core and 2,831 in Multi-Core.
As our fellow Applesfera colleagues explain, this is all the more remarkable when you consider that the DTK is running Geekbench in emulation via Rosetta 2 with a two-year-old processor running at a lower frequency (2.4 GHz vs. 3 GHz in the core of the Snapdragon SQ1) and which is theoretically at a disadvantage.
Doing a little math on the A12Z benchmarks vs the DTK, Rosetta is running at about 70% of native performance - on an iPad chip that's basically two years old, not designed for macOS. I imagine Geekbench surfaces both static and JIT translation, so static-only apps may be faster- Steve Troughton-Smith (@stroughtonsmith) June 29, 2020
Developer Steve Throughton-Smith indicated that according to their accounts Rosetta 2 allowed the benchmark to run with roughly 70% of what the performance would be in the native application. That also has merit because the chip is as we say already veteran and was not designed to run macOS. In fact, other developers point out that it is only using four of the eight cores that the A12Z has (the four most powerful, yes).
All this suggests that the room for maneuver for Apple here is remarkable, and of course the performance of these "raw" teams is promising for future Apple teams that have dedicated chips and that theoretically will be significantly higher in power as they are designed not for mobile phones or tablets, but for laptops or all-in-one computers like the iMac.
What happens when comparing with Intel machines?
The truth is that comparing that Developer Transition Kit with the Apple A12Z with a Surface Pro X may not be so striking when precisely what is intended is to know how future Mac ARMs will behave when compared to equivalent machines based on microphones from Intel or AMD .
At the moment there are no direct tests that allow this comparison, but we do have, for example, the analysis that SixColors made of the MacBook Air with the Core i5 in its most recent edition of 2020.
When comparing the performance of these devices with that of the 2018 iPad Pro, the results were also surprising in favor of Apple's ARM chips. In the Geekbench 5 test the iPad Pro outperformed the MacBook Air in both single-core and multi-core performance.
The chip of the iPad Pro (2018) is in fact the Apple A12X, while the Apple A12Z used in those Apple DTKs is according to experts a "refined" version of the previous one with one more core for its GPU.
It should be clarified that in that comparison the iPad Pro ran the version of Geekbench for iOS and its ARM micros while the MacBook Air ran the version for macOS and its Intel micro with AMD64 / x86-64 architecture. Compared to the Surface Pro X that "Mac mini ARM" ran Geekbench for Intel via emulation, we insist, something that puts it at an even greater disadvantage when checking data.
It remains to be seen how far Apple can go with those future Macs based on "Apple Silicon", but of course these first data -and everything we've seen before- makes expectations are remarkable for those teams.